Re: [recommenders-dev] Need help to collect completion proposals made by

[Andreas Sewe <andreas.sewe@xxxxxxxxxxxxxx>]

Hi Muhammad,

I may not be able to answer all of your questions, but hopefully some of
them.

> The program generated result is different than the completion
> proposals proposed by the code recommender plug-in (for the later case
> I write down the code and invoke method completion manually). The
> version of code recommenders source code I am using is 1.0 whereas the
> plug-in version number is 2.0.
> The model file I am using is jre-1.0.0-call.zip in my program.
> 
> Could you please take a look on my following questions and let me know
> your comment/suggestions:
> 
> i) How can I obtain the new model for the swing, AWT and java.io
> library? Among these three I have only model for swing (but may be the
> model is also old).

The version of the model file (1.0.0) is misleading; it does not mean
that the model's version is 1.0.0, but rather that jre-1.0.0-call.zip is
the call model for the JRE in version 1.x.y. (The zeros really mean "any
minor and micro version").

That being said, there is now (as of Code Recommenders 2.0) something
akin to the version of the model itself, namely a timestamp after the
1.0.0:
<http://download.eclipse.org/recommenders/models/luna-m3/jre/jre/1.0.0-SNAPSHOT/?d>

Also, we take care to use new model repository URLs whenever we mine new
models (the above JRE model is from our "luna-m3" release). What model
repository are you using in your measurements? And what mode repository
have you configured in your plugin preferences [1]?

> ii) Do you think the result is different because I am using old
> version of code recommenders or old model file or my calculation is
> wrong?.

I would definitely advise you to use the Code Recommenders 2.0 code base
for your experiments. The old 1.0 branch was very hard to use in a
headless environment like the one your want for your evaluation. Also,
there may be fewer bugs in 2.0 than in 1.0.

> iii) Can you also provide me any hint about the number of examples
> used to build the model for Swing/AWT and SWT library.

I don't have exact numbers, but for the most common types (String, List,
and friends) we have more than 200k individual usages (but we cut of at
250.000). For other types, there are less usages. If you provide me with
a list of frameworks you are interested in, I can provide you with
counts *per type*.

Or do you mean by "number of examples" the number of JARs depending on,
say, SWT that went into the data-miner?

Hope this helps.

Andreas

[1] <http://www.eclipse.org/recommenders/manual/#models-preferences>

-- 
Codetrails UG (haftungsbeschränkt)
The knowledge transfer company

Robert-Bosch-Str. 7, 64293 Darmstadt
Mobile: +49-170-811-3791
http://www.codetrails.com/

Managing Director: Dr. Marcel Bruch
Handelsregister: Darmstadt HRB 91940